Traffic circles in South Africa : traffic performance and driver behaviour.

Abstract

This thesis presents the results of an investigation into traffic operations and driver behaviour at traffic circles under South African conditions. The scarcity of local traffic circles necessitated the development of a simulation program (TRACSIM) to assist in the research process. This microscopic program for single lane circles is based on event updates and was calibrated and validated based on local data. Because the acceptance of gaps is such a vital part of the operation of a traffic circle, it was examined in detail. Specific attention was given to the possible use of a gap acceptance model based on variables other than time. Since the gap acceptance process also depends on the gap distribution in the circulating stream, the effect of the origin-destination pattern was also investigated. Two existing analysis techniques are evaluated and verified for local conditions, improving them where possible. Generally these techniques under-estimate traffic delay at local circles. Observations indicate a difference between the acceptance of gaps/lags in the entering and circulating stream of conflicting traffic as well as a difference between critical gaps and critical lags. The mean observed critical gaps/lags are larger than in other countries, which indicates that delays at local circles will be greater. Gap/lag acceptance based on critical distances rather than critical times was applied successfully in the simulation program TRACSIM. A method is proposed to estimate critical distances from the geometric layout of the circle. Critical gaps are not fixed, but should vary with at least the conflicting flows. The investigation of the effect of unbalanced flows on delay, showed that the variability in drivers' critical gaps is more a function of delay than of conflicting flow. Entry delays increase because of an increase in conflicting flows or because of an unfavourable imbalance of conflicting flows. In both instances the drivers' critical gaps will decrease. A variable critical gap model only based on conflicting flows will show no change in the drivers' critical gaps if the conflicting volumes remain constant, even though the actual average delay might increase because of an unfavourable imbalance in conflicting flows.